Build on Salesforce with your AI agent
Hutte now works headless over the Model Context Protocol. Connect Claude, Cursor or any MCP client and let your AI assistant develop in a real Salesforce org β then review, commit and ship, with you in control.
Implement HUT-142 in a Sandbox.
On it. Pulling the feature from Hutteβ¦
β Loaded requirements for HUT-142
β Connected to Sandbox dev-142
β Updated Apex & flow metadata
Ready for review β 6 components changed.
Looks good β commit & open a PR.
Salesforce DevOps, now agent-native
Hutte exposes your projects, features and orgs through the Model Context Protocol β so your AI assistant can do the work, while you stay in control.
Your AI, your workflow
Works with Claude, Claude Code, Cursor and any MCP-compatible client. Drive Hutte from the AI tool your team already uses β no context switching.
Real orgs, on demand
Hutte provisions and connects a genuine Salesforce environment β Sandbox or Scratch Org β and hands the secure connection to your agent so it can build against live metadata.
You stay in control
Inspect exactly what changed as a clear delta of XML metadata, test the result in Hutte, then commit and open a pull request β from Hutte or straight from your AI tool.
A typical agentic workflow
From requirement to pull request β your AI does the heavy lifting, Hutte keeps it safe and reviewable.
Start with a feature
Create the feature in Hutte, from your Jira issue, or just ask the agent to set it up β it doesn't matter where you begin. The feature is the glue that ties your requirements, environment and changes together, wherever those requirements actually live.
Ask your AI to implement it
In Claude, Cursor or your favourite MCP client, simply instruct the assistant to implement the feature. No copy-pasting context, no manual setup.
The agent builds in a real org
Your assistant pulls the requirements, uses the secure org connection Hutte provides, and implements the changes against a live Salesforce environment.
Review, test and inspect changes
Test the work in Hutte and inspect exactly what changed as a delta of XML metadata β so nothing ships that you haven't seen.
Commit and open a pull request
Commit and open a PR from Hutte, or do it straight from your AI tool. Either way, your changes land in Git with a clean, reviewable history.
Deploy with confidence
Once the PR is merged, promote the change through your environments β from QA all the way to production β using Hutte's deployment workflows, the same way you ship every other change.
Works with both Sandboxes and Scratch Orgs
Whether your team develops in Sandboxes or Scratch Orgs, Hutte gives your agent a real, secure environment to build in β including pooled orgs that are ready the moment you need them.
Frequently Asked Questions
What is the Hutte MCP?
The Hutte MCP is a Model Context Protocol server that connects AI assistants β like Claude, Claude Code and Cursor β directly to your Hutte projects, features and orgs. It lets you run agentic Salesforce DevOps workflows in natural language, without leaving your AI tool. See the documentation for setup details.
Does it work with Sandboxes and Scratch Orgs?
Yes. Hutte connects your agent to a real Salesforce environment β it works with both Sandboxes and Scratch Orgs, including pooled orgs that are ready to use instantly.
Which AI tools are supported?
Any MCP-compatible client. That includes Claude.ai, Claude Code and Cursor today, with more clients supported as the MCP ecosystem grows.
Do I still review changes before they ship?
Absolutely. You can test the agent's work in Hutte and inspect every change as a delta of XML metadata before you commit. Pull requests can be opened from Hutte or directly from your AI tool β you decide what gets merged.
Bring your AI agent to Salesforce DevOps
Connect your favourite AI tool to Hutte and let it develop in real orgs β while you keep full control over what ships.